Hadoop Virtualization Courtney Webster Make Data Work strataconf.com Presented by O’Reilly and Cloudera, Strata + Hadoop World is where cutting-edge data science and new business fundamentals intersect— and merge n n n Learn business applications of data technologies Develop new skills through trainings and in-depth tutorials Connect with an international community of thousands who work with data Job # 15420 Hadoop Virtualization Courtney Webster Hadoop Virtualization by Courtney Webster Copyright © 2015 O’Reilly Media, Inc All rights reserved Printed in the United States of America Published by O’Reilly Media, Inc., 1005 Gravenstein Highway North, Sebastopol, CA 95472 O’Reilly books may be purchased for educational, business, or sales promotional use Online editions are also available for most titles (http://safaribooksonline.com) For more information, contact our corporate/institutional sales department: 800-998-9938 or corporate@oreilly.com Editors: Julie Steele and Jenn Webb February 2015: Illustrator: Rebecca Demarest First Edition Revision History for the First Edition: 2015-01-26: First release 2015-03-16: Second release The O’Reilly logo is a registered trademark of O’Reilly Media, Inc Hadoop Virtuali‐ zation, the cover image, and related trade dress are trademarks of O’Reilly Media, Inc While the publisher and the author have used good faith efforts to ensure that the information and instructions contained in this work are accurate, the publisher and the author disclaim all responsibility for errors or omissions, including without limi‐ tation responsibility for damages resulting from the use of or reliance on this work Use of the information and instructions contained in this work is at your own risk If any code samples or other technology this work contains or describes is subject to open source licenses or the intellectual property rights of others, it is your responsibility to ensure that your use thereof complies with such licenses and/or rights ISBN: 978-1-491-90676-7 [LSI] Table of Contents The Benefits of Deploying Hadoop in a Private Cloud Abstract Introduction MapReduce Hadoop Virtualizing Hadoop Another Form of Virtualization: Aggregation Benefits of Hadoop in a Private Cloud Agility and Operational Simplicity with Competitive Performance Improved Efficiency Flexibility Conclusions Apply the Resources and Best Practices You Already Know Benefits of Virtualizing Hadoop 2 10 12 15 15 16 iii The Benefits of Deploying Hadoop in a Private Cloud Abstract Hadoop is a popular framework used for nimble, cost-effective anal‐ ysis of unstructured data The global Hadoop market, valued at $1.5 billion in 2012, is estimated to reach $50 billion by 2020.1 Companies can now choose to deploy a Hadoop cluster in a physical server envi‐ ronment, a private cloud environment, or in the public cloud We have yet to see which deployment model will predominate during this growth period; however, the security and granular control offered by private clouds may lead this model to dominate for medium to large enterprises When compared to other deployment models, a private cloud Hadoop cluster offers unique benefits: • A cluster can be set up in minutes • It can flexibly use a variety of hardware (DAS, SAN, NAS) • It is cost effective (lower capital expenses than physical deploy‐ ment and lower operating expenses than public cloud deploy‐ ment) • Streamlined management tools lower the complexity of initial configuration and maintenance • High availability and fault tolerance increase uptime This report reviews the benefits of running Hadoop on a virtualized or aggregated (container-based) private cloud and provides an over‐ view of best practices to maximize performance Introduction Today, we are capable of collecting more data (and various forms of data) than ever before.2 It may be the most valuable intangible asset of our time The sheer volume (“big data”) and need for flexible, lowlatency analysis can overwhelm traditional management systems like structured relational databases As a result, new tools have emerged to store and mine large collections of unstructured data MapReduce In 2004, Google engineers described a scalable programming model for processing large, distributed datasets.3 This model, MapReduce, abstracts computation away from more complicated tasks like data distribution, failure handling, and parallelization Developers specify a processing (“map”) function that behaves as an independent, mod‐ ular operation on blocks of local data The resulting analyses can then be consolidated (or “reduced”) to provide an aggregate result This model of local computation is particularly useful for big data, where the transfer time required to move the data to a centralized computing module is limiting Hadoop Doug Cutting and others at Yahoo! combined the computational pow‐ er of MapReduce with a distributed filesystem prototyped by Google in 2003.4 This evolved into Hadoop—an open source system made of MapReduce and the Hadoop Distributed File System (HDFS) HDFS makes several replica copies of the data blocks for resilience against server failure and is best used on high I/O bandwidth storage devices In Hadoop 1.0, two master roles (the JobTracker and the Namenode) direct MapReduce and HDFS, respectively Hadoop was originally built to use local data storage on a dedicated group of commodity hardware In a Hadoop cluster, each server is considered a node A “master” node stores either the JobTracker of MapReduce or the Namenode of HDFS (although in a small cluster as shown in Figure 1, one master node could store both) The remaining servers (“worker” nodes) store blocks of data and run local computa‐ tion on that data | The Benefits of Deploying Hadoop in a Private Cloud Figure A simplified overview of Hadoop5 The JobTracker directs low-latency, high-bandwidth computational jobs (TaskTrackers) on local data The Namenode, the lead storage directory of HDFS, provides rack awareness: the system’s knowledge of where files (Data Nodes) are stored among the array of workers It does this by mapping HDFS file names to their constituent data blocks, and then further maps those data blocks to Data Node processes This knowledge is responsible for HDFS’s reliability, as it ensures nonredundant locations of data replicates Hadoop 2.0 In the newest version of Hadoop, the JobTracker is no longer solely reponsible for managing the MapReduce programming framework The JobTracker function is distributed, among others, to a new Ha‐ doop component called the Application Master In order to run tasks, ApplicationMasters request resources from a central scheduler called the ResourceManager This architectural redesign improves scalability and efficiency, bypassing some of the limitations in Hadoop 1.0 A new central scheduler, the ResourceManager, acts as its key replacement Developers can then construct ApplicationMasters to encapsulate any knowledge of the programming framework, such as MapReduce In The Benefits of Deploying Hadoop in a Private Cloud | order to run their tasks, ApplicationMasters request resources from the ResourceManager This architectural redesign improves scalability and efficiency, bypassing some of the limitations in Hadoop 1.0 Virtualizing Hadoop As physically deployed Hadoop clusters grew in size, developers asked a familiar question: can we virtualize it? Like other enterprise (and Java-based) applications, development ef‐ forts moved to virtualization as Hadoop matured A virtualized private cloud uses a group of hardware on the same hypervisor (such as vSphere [by VMware], XenServer [by Citrix], KVM [by Red Hat], or Hyper-V [by Microsoft]) Instead of individual servers, nodes are vir‐ tual machines (VMs) designated with master or worker roles Each VM is allocated specific computing and storage resources from the physical host, and as a result, one can consolidate their Hadoop cluster onto far fewer physical servers There is an up-front cost for virtuali‐ zation licenses and supported or enterprise-level software, but this can be offset with the cluster’s decreased operating expenses over time Virtualizing Hadoop created the infrastructure required to run Ha‐ doop in the cloud, leading major players to offer web-service Hadoop The first, Amazon Web Services, began beta testing their Elastic Map‐ Reduce service as early as 2009 Though public cloud deployment is not the focus of this review, it’s worth noting that it can be useful for ad hoc or batch processing, especially if your data is already stored in the cloud For a stable, live cluster, a company might find that building its own private cloud is more cost effective Additionally, regulated industries may prefer the security of a private hosting facility In 2012, VMware released Project Serengeti—an open source man‐ agement and deployment platform on vSphere for private cloud en‐ vironments Soon thereafter, they released Big Data Extensions (BDE), the advanced commercial version of Project Serengeti (run on vSphere Enterprise Edition) Other offerings, like OpenStack’s Project Sahara on KVM (formerly called Project Savanna), were also released in the past two years | The Benefits of Deploying Hadoop in a Private Cloud Though these programs run on vendor-specific virtualization plat‐ forms, they support most (if not all) Hadoop distributions (Apache Hadoop [1.x and 2.x] and commercial distributions like Cloudera, Hortonworks, MapR, and Pivotal) They can also manage coordinat‐ ing applications (like Hive and Pig) that are typically built on top of a Hadoop cluster to satisfy analytical needs Case Study: Hadoop on a Public Versus Private Cloud A company providing enterprise business solutions initially turned to the public cloud for its analytics applications Ad hoc use of a Ha‐ doop cluster of 200 VMs cost about $40k a month When their de‐ velopers needed consistent access to Hadoop, the bills would spike by an additional $20-40k For $80k, they decided to build their own 225 TB, 30-node virtualized Hadoop cluster Flash-based SAN and serverbased flash cards were used to enhance performance for 2-3 TB of very active data Using Project Serengeti, it took about 10 minutes to deploy their cluster Another Form of Virtualization: Aggregation Cloud computing without virtualization Thus far, virtualization refers to using a hypervisor and VMs to isolate and allocate resources in a private cloud environment For clarity, “virtualization” will continue to be used in this context But building a private cloud environment isn’t limited to virtualization Aggrega‐ tion (as a complement to or on top of virtualization) became a useful alternative for cloud computing (see B in Figure 2), especially as ap‐ plications like Hadoop grew in size The Benefits of Deploying Hadoop in a Private Cloud | Figure Strategies for cloud computing Virtualization partitions servers into isolated virtual machines, while aggregation consolidates servers to create a common pool of resources (like CPU, RAM, and memory) that applications can share System containers can run a full OS, like a VM, while others (application con‐ tainers) contain a single process or application This allows multiple applications to access the consolidated resources without interfering with each other Resources can be dynamically allocated to different applications as their loads change In an initial study by IBM, Linux containers (LXC) and control groups (cgroups) allowed for isolation and resource control in an aggregated environment with less overhead than a KVM hypervisor.6 The poten‐ tial overhead advantages should be weighed against some limitations with LXC, such as the restriction to only run on Linux and that, cur‐ rently, containers offer less performance isolation than VMs | The Benefits of Deploying Hadoop in a Private Cloud If an industry has already invested in virtualization licenses, aggrega‐ tion can be used on a virtualized environment to provide one “super” VM (see C in Figure 2) Unless otherwise specified, however, the terms “aggregation” and “containers” here imply use on a bare metal (nonvirtualized) environment Cluster managers Cluster managers and management tools work on the application level to manage containers and schedule tasks in an aggregated environ‐ ment Many cluster managers, like Apache Mesos (backed by Meso‐ sphere) and StackIQ, are designed to support analytics (like Hadoop) alongside other services Hadoop on Mesos Apache Mesos provides a foundational layer for running a variety of distributed applications by pooling resources Mesos allows a cluster to elastically provision resources to multiple applications (including more than one Hadoop cluster) Mesosphere aims to commercialize Mesos for Hadoop and other enterprise applications while building add-on frameworks (like distributed schedulers) along the way In 2013, they released Elastic Mesos to easily provision a Mesos cluster on Amazon Web Services, allowing companies to run Hadoop 1.0 on Mesos in bare-metal, virtualized, and now public cloud environ‐ ments Benefits of Hadoop in a Private Cloud In addition to cost-effective setup and operation, private cloud de‐ ployment offers additive value by streamlining maintenance, increas‐ ing hardware utilization, and providing configurational flexibility to enhance the performance of a cluster The Benefits of Deploying Hadoop in a Private Cloud | Agility and Operational Simplicity with Competitive Performance Deploy a Scalable, High-Performance Cluster with a Simplified Management Interface • Benchmarking tools indicate that the performance of a virtual cluster is comparable to a physical cluster • Built-in workflows lower initial configuration complexity and time to deployment • Streamlined monitoring consoles provide quick performance read-outs and easy-to-use management tools • Nodes can be easily added and removed for facile scaling Competitive performance Since a hypervisor demands some amount of computational resour‐ ces, initial concerns about virtual Hadoop focused on performance The virtualization layer requires some CPU, memory, and other re‐ sources in order to manage its hosted VMs,7 though the impact is dependent on the characteristics of the hypervisor used Over the past to 10 years, however, the performance of VMs have significantly improved (especially for Java-based applications) Many independent reports show that when using best practices, a vir‐ tual Hadoop cluster performs competitively to a physical system.8,9 Increasing the number of VMs per host can even lead to enhanced performance (up to 13%) Container-based clusters (like Linux VServer, OpenVZ, and LXC) can also provide near-native performance on Hadoop benchmarking tests like WordCount and TeraSuite.10 With such results, performance concerns are generally outweighed by the numerous other benefits provided by a private-cloud deployment Rapid deployment To deploy a cluster, Hadoop administrators must navigate a compli‐ cated setup and configuration procedure Clusters can be composed | The Benefits of Deploying Hadoop in a Private Cloud of tens to hundreds of nodes—in a physical deployment, each node must be individually configured With a virtualized cluster, an administrator can speed up initial con‐ figuration by cloning worker VM nodes VMs can be easily copied to expand the size of the cluster, and problematic nodes can be removed and then restored from backup images Some virtualized Hadoop of‐ ferings, like BDE, can entirely automate installation and network con‐ figuration Using containers instead of VMs offers deployment advantages as well, as it takes hours to provision bare metal, minutes to provision VMs, but just seconds to provision containers Like BDE, cluster managers can also automate installation and configuration (including network‐ ing software, OS software, and hardware parameters, among others) Improved management and monitoring A Hadoop cluster must be carefully monitored to meet the demands of 24/7 accessibility, and a variety of management tools exist to help watch the cluster Some come with your Hadoop distribution (like Cloudera Manager and Pivotal’s Command Center), while others are open source (like Apache Ambari) or commercial (like Zettaset Or‐ chestrator) Virtualization-aware customers are already using hyper‐ visor management interfaces (like vCenter or XenCenter) to simplify resource and lifecycle management, and a virtualized Hadoop cluster integrates as just another monitored workload These simplified provisioning and management tools enable Hadoopas-a-service Some platforms allow an administrator to hand off preconfigured templates, leaving users to customize the environment to suit their individual needs More sophisticated cloud management tools automate the deployment and management of Hadoop, so com‐ panies can offer Hadoop clusters without users managing any config‐ urational details Scalability Modifying a physical cluster—removing or adding physical nodes— requires a reshuffling of the data within the entire system Load bal‐ ancing (ensuring that all worker nodes store approximately the same amount of data) is one of the most important tasks when scaling and maintaining a cluster Some hypervisors, like vSphere Enterprise Ed‐ The Benefits of Deploying Hadoop in a Private Cloud | ition, include distributed resource schedulers that can perform auto‐ matic load balancing To scale an aggregated system, cluster managers just need to be in‐ stalled on new nodes When the cluster scheduler is made aware of the new node, it will automatically absorb the offered resources and begin scheduling tasks on it Improved Efficiency Create a Robust, High-Utilization Cluster • Rather than monopolizing dedicated hardware, a private cloud cluster allows for mixed workflows for higher utilization • High availability and fault tolerance increase the uptime of a cluster during unanticipated outages and failures or routine maintenance Higher resource utilization A physical deployment model monopolizes its dedicated hardware Physical Hadoop clusters are often over-engineered—they are built to handle an estimated peak capacity, but left underutilized the rest of the time Any complementary application (like a NoSQL or SQL da‐ tabase) requires its own dedicated hardware as well In a virtual deployment, resources like CPU and RAM are partitioned for the Hadoop cluster, freeing up resting resources for other tasks Co-locating VMs running Hadoop roles (like MapReduce jobs) with VMs running other workloads (such as Hive queries on HBase) can balance the use of a system.5 Multiple workloads can be run concur‐ rently on the same hardware with a minimal effect on results (less than a 10% difference when compared to utilizing separate, independent workloads on a standalone cluster).11 An aggregated cloud also offers higher utilization Though isolated from one another, all applications access the same pool of resources The system can elastically scale resources for each application Theo‐ retically, a high-load application could use the entirety of aggregated resources (like CPU, RAM, and memory) until loads on other appli‐ cations increase 10 | The Benefits of Deploying Hadoop in a Private Cloud Minimizing downtime with high availability and fault tolerance High availability (HA) protects a cluster during planned and unplan‐ ned downtime Failovers can be deliberately triggered for maintenance or are automatically triggered in the event of failures or unresponsive service Virtualized HA solutions monitor hosts and VMs to detect hardware and guest operating system (OS) failures If a server outage or failed network connection is detected, VMs from the failed host are restarted on new hosts without manual intervention (see Figure 3).12 In the case of an OS failure, VMs are automatically restarted In aggregated en‐ vironments, failed workloads automatically failover to a new node with available resources HA in a Hadoop cluster can protect against the single-point failure of a master node (the Namenode or JobTracker) If desired, the entire cluster (master nodes and worker nodes) can be uniformly managed and configured for HA.12 Figure High availability monitoring12 In a virtualized environment, fault tolerance (FT) provides continuous availability by creating a live, up-to-date shadow (a secondary in‐ stance) of a VM Though an FT system is not able to detect if the application fails or a guest OS hangs,13 it triggers a failover procedure to the secondary VM if a VM stops due to a hardware outage or loss of network connectivity This helps prevent data loss and decreases The Benefits of Deploying Hadoop in a Private Cloud | 11 downtime Combining HA and FT can create maximum availability for a virtualized Hadoop cluster Flexibility Utilize Options for Flexible Configuration • A cluster can be built using DAS, SAN/NAS, or a hybrid combi‐ nation of storage • Configurational options create elastic scalability to address fluc‐ tuating demands Hardware flexibility By using commodity hardware and built-in failure protection, Ha‐ doop was designed for flexibility Virtualization takes this a step fur‐ ther by abstracting away from hardware completely A private cloud can use direct attached storage (DAS), a storage attached network (SAN), or a network attached storage (NAS) SAN/NAS storage can be more costly but offers enhanced scalability, performance, and data isolation If a company has already invested in non-local storage, their Hadoop cluster can strategically employ both direct- and networkattached devices The storage or VMDK files for the Namenode and JobTracker can be placed on SAN for maximum reliability (as they are memory- but not storage-intensive) while worker nodes store their data on DAS.5 The temporary data generated during MapReduce can be stored wherever I/O bandwidth is maximized Case Study: Hadoop on NAS A major shipping company uses a virtualized Hadoop cluster to per‐ form web log analysis (detecting mobile devices accessing the web‐ site), ZIP code analysis (which ZIP codes are the highest source or destination for shipments), and shipment analysis (to determine pat‐ terns that may delay a package) Their cluster is hosted on EMC Isilon NAS storage From their perspective, Isilon helps drive down the total cost of ownership by eliminating the “triple replicate penalty” with data storage Additionally, they’ve found that a fast enough network 12 | The Benefits of Deploying Hadoop in a Private Cloud can perform competitively to DAS (equalizing the playing field in terms of data locality) Configurational flexibility As previously described, organizing computation tasks to run on blocks of local data (data locality) is the key to Hadoop’s performance In a physical deployment, this necessitates that worker nodes host data and compute roles in a fixed 1:1 ratio (a “combined” model) For this model to be mimicked in a virtual cluster, each hypervisor server would host one or more VMs that contained data and compute pro‐ cesses (see A in Figure 4) These configurations are valid, but difficult to scale under practical circumstances Since each VM stores data, the ease of adding or removing nodes (simply copying from a template or using live migrate capabilities) would be offset by the need to rebalance the cluster If instead, compute and data roles on the same hypervisor server were in separate VMs (see B in Figure 4), compute operations could be scaled according to demand without redistributing any data.14 Like‐ wise in an aggregated cloud, Apache Mesos only spins up TaskTracker nodes as a job runs When the task is complete, the TaskTrackers are killed, and their capacity is placed back in the consolidated pool of resources This “separated” model is fairly intuitive (since the TaskTracker con‐ trols MapReduce and the Namenode controls the data storage) and the flexibility of virtualized or aggregated clusters makes it relatively simple to configure In addition to creating this elastic system (where compute processes can be easily cloned or launched to increase throughput), the separated model also allows you to build a multi-tenant system where multiple, isolated compute clusters can operate on the same data The same cloud could host a production Hadoop cluster as well as a development and QA environment The Benefits of Deploying Hadoop in a Private Cloud | 13 Figure Configurational flexibility with compute and data processes (C: compute; D: data) Figure modified from VMware’s “Deploying Virtualized Hadoop Systems with VMware vSphere Big Data Exten‐ sions (BDE).”5 Virtual rack awareness Using the separated model does carry an added complication If com‐ pute processes can scale on demand, the cluster’s topology is dynamic (compared to the fixed structure of a physical deployment) In a virtualized cluster, a compute and data node on the same hyper‐ visor server communicate over a virtual network in memory (without suffering physical network latencies) It is important that the system maintain virtual “rack awareness” to preserve the performance ad‐ vantage of data locality To accommodate this need, VMware contrib‐ uted a tool called Hadoop Virtualization Extensions (HVE) to Ha‐ doop.15 HVE accounts for the virtualization layer by grouping all VMs on the same host in a new domain For best performance, the cluster can use this domain to direct computation to perform on the same hypervisor server hosting the data It can also intelligently place data replicates on separate hosts to provide maximum protection in the event of a hardware failure 14 | The Benefits of Deploying Hadoop in a Private Cloud Separately scaled data and compute processes presents a similar chal‐ lenge in an aggregated environment If an application can request tasks and be offered resources throughout the entire cloud, how does it know which nodes store the data required? Adding constraints allows an application to reject resource offers that don’t meet its require‐ ments.16 A technique called “delay scheduling,” in which an application can wait if it cannot launch a local task, can result in nearly optimal data locality.17 Conclusions Apply the Resources and Best Practices You Already Know If planning your first cluster or a deployment overhaul, it is important to consider the following: • Current data storage • Estimated data growth • The amount of temporary data that will be stored during Map‐ Reduce processing • Throughput and bandwidth needs • Performance needs • The resources (hardware and software) you have available to ded‐ icate to the cluster • The resources (hardware and software) you’d need to purchase to dedicate to the cluster It would be difficult to specify an ideal architecture for every Hadoop cluster, as analytical demands and resource needs vary widely But planning a private cloud cluster doesn’t necessitate a large learning curve either Many companies can utilize resources (e.g., virtualization licenses, cluster managers, DAS, or SAN/NAS storage) they already have Ad‐ ditionally, many of the best practices an IT department already puts in place (like avoiding VM contention and optimizing I/O bandwidth) translate well to configuring a high-performance Hadoop cluster The Benefits of Deploying Hadoop in a Private Cloud | 15 Benefits of Virtualizing Hadoop Whether Hadoop is deployed in a physical system or in a private or public cloud, the goals of a well-established infrastructure are the same In any implementation, Hadoop should provide: • Cost-effective and high-performance data analysis • Failure-tolerant data storage • Scalable capability for future growth • Minimal downtime For Hadoop administrators and users, a private cloud offers unique benefits with comparable (even improved) performance Rapid de‐ ployment and built-in workflows ease initial configuration complex‐ ity, to the point where developers can use built-in tools to configure their own test environments without involving IT Management tools make it easier to monitor and analyze performance, and features like high availability and fault tolerance decrease downtime The need for low-latency data management systems will grow in com‐ ing years, and enterprise applications continue to move from physical systems into cloud computing infrastructures Using a private cloud can create a scalable, streamlined Hadoop cluster built to accommo‐ date data science’s evolving landscape 16 | The Benefits of Deploying Hadoop in a Private Cloud Notes “Global Hadoop Market (Hardware, Software, Services, HaaS, End User Application and Geography) - Industry Growth, Size, Share, In‐ sights, Analysis, Research, Report, Opportunities, Trends and Fore‐ casts Through 2020” “Big Data: The Next Big Thing.” Nasscom, New Delhi, 2012 “MapReduce: Simplified Data Processing on Large Clusters” Hadoop: The Definitive Guide VMware, Inc “Deploying Virtualized Hadoop Systems with VMware vSphere Big Data Extensions.” 2014 Felter, W., A Ferreira, R Rajamony, and J Rubio “An Updated Per‐ formance Comparison of Virtual Machines and Linux Containers.” IBM Research Report, 2014 Microsoft IT Big Data Program “Performance of Hadoop in HyperV.” MSIT SES Enterprise Data Architect Team, 2013 Buell, J “A Benchmarking Case Study of Virtualized Hadoop Per‐ formance on VMware vSphere 5.” Buell, J “Virtualized Hadoop Performance with VMware vSphere 5.1.” VMware, Inc., 2013 10 Xavier, M.G., M.V Neves, and A.F De Rose “Performance Com‐ parison of Container-based Virtualization MapReduce Clusters.” IEEE, 2014 11 Intel, VMware, and Dell “Scaling the Deployment of Multiple Ha‐ doop Workloads on a Virtualized Infrastructure.” 2013 12 Hortonworks and VMware “Apache Hadoop 1.0 High Availability Solution on VMware vSphere.” 2011 13 Buell, J “Protecting Hadoop with VMware vSphere Fault Toler‐ ance.” VMware, Inc., 2012 14 Magdon-Ismail, T., et al “Toward an Elastic Elephant – Enabling Hadoop for the Cloud.” VMware Technical Journal 2, no (December 2013): 56-64 The Benefits of Deploying Hadoop in a Private Cloud | 17 15 VMware, Inc “Hadoop Virtualization Extensions on VMware vSphere 5.” 16 B Hindman et al “Mesos: A Platform for Fine-Grained Resource Sharing in the Data Center.” UC Berkeley, 2010 17 M Zaharia et al “Delay Scheduling: A Simple Technique for Ach‐ ieving Locality and Fairness in Cluster Scheduling.” Proceedings of the 5th European Conference on Computer Systems, 2010 18 | The Benefits of Deploying Hadoop in a Private Cloud About the Author Courtney Webster is a freelance writer with professional experience in laboratory automation, automated data analysis, and the applica‐ tion of mobile technology to clinical research You can follow her on Twitter @automorphyc and find her blog at http://automorphyc.com  ... Virtualization: Aggregation Benefits of Hadoop in a Private Cloud Agility and Operational Simplicity with Competitive Performance Improved Efficiency Flexibility Conclusions Apply the Resources... configurational flexibility to enhance the performance of a cluster The Benefits of Deploying Hadoop in a Private Cloud | Agility and Operational Simplicity with Competitive Performance Deploy... initial configuration complex‐ ity, to the point where developers can use built-in tools to configure their own test environments without involving IT Management tools make it easier to monitor